Continual Development of a Personalized Decision Support System

Continual Development of a Personalized Decision Support System Dina Demner-Fushman Charlotte Seckman Cheryl Fisher George Thoma

Background Collaboration between: National Institutes of Health (NIH) • Nursing and Patient Care Services • Dept. of Clinical Research Informatics National Library of Medicine (NLM) NIH Clinical Research Center National Library of Medicine

Purpose • Provide access to key facts needed to support clinical decision making and evidence based plans of care Approach • Design and build a system which enables a medical institution to automatically augment a patient’s Electronic Medical Record (EMR) with pertinent patient-specific information from NLM's and other evidence-based resources 3

Info. Bot Architecture and Processes NIH Clinical Center NLM EBP Info. Bot CRIS Database 2 5 3 Send HTTP request to Info. Bot 7 Show Info. Bot dashboard Click Info. Bot tab Find problems, interventions, anatomic locations in text 2. Split medication lists, find names 3. Look-up rules for specific problems, compose searches 4. Get information 5. Fill out the dashboard template 6. Send response 4 Get stored procedures for retrieval of predefined CRIS fields 1 1. 6

Continuous evaluation • Prototype evaluation • Focus Group – Barriers – Enhancers • Access and Usage Data – 6 months & ongoing post implementation • Online Survey Questionnaire – User Satisfaction – Ease of Navigation 5

Prototype evaluation 6

Evaluation Questions • To what extent is the Info. Bot output relevant to the patient cases presented? • To what extent is the Info. Bot output useful in clinical care? • How easy is the Info. Bot system to use? • How often would the Info. Bot system be used?

Conclusions • Info. Bot system is easy to use • 75% respondents would use it at least once a week • 56. 2% respondents indicated a need for more information • Overall document types were found to be relevant and useful in the majority of cases 8

Recommendations • Continue development of Info. Bot system to address evaluator’s feedback (both resources and functions) • Improve MEDLINE retrieval quality • Develop the Info. Bot 2. 0 interface • Integrate local resources • Evaluate the interface using focus groups • Expand user base • Integrate with CRIS

Project Needs • Staffing: – Continued project leader commitment from NLM, DCRI, & NPCS – Clinical staff to provide feedback • Resources: – More information about investigational drugs and procedures, protocols – Unit guidelines, standards of care – Care plan text book, Mosby's Nursing materials – Micromedex – More patient education materials – FDA recalls, market withdrawals and safety alerts

Info. Bot Dashboard I

Sunrise Clinical Manager

Access and Usage • 1545 unique users (six months) • 861 returning users • Between 2 and 4 hits an hour covering all hours Link Access Number of Hits Clinical Trials 930 Medications 1132 Medline Plus 293 Definitions 221 Standards of Practice/Mosby Pub. Med Articles 25 407 (290 votes)

Focus Groups Barriers to using Info. Bot • Time Constraints – Search articles but no time to read them – Short time span to capture relevant information – Time to do reading not incorporated into the day • Limited Training – Not reaching nurses that do not attend NPC or CPC meetings – Reading an abstract or interpreting the research • Relatedness to patient – Managing chief complaint – Pertinent diagnosis information – Patients who are on multiple protocols

Focus Groups • Enhancers – Access to medication information – Easy to use and quick retrieval – Information condensed in one location • Recommendations – – – Incorporate EBP Info. Bot training into CRIS orientation Develop drop down lists to enter problems Create fact sheets Allow time for Nursing Rounds to review articles Picture of pills Sample discharge planning

Info. Bot dashboard III

Challenges • Technical trade-offs – Real time processing vs. quality • NLP parsers are slow • Best evidence searches are iterative • Continuous collaboration – EHR format and storage change • Need notifications • Time for system update • Users’ time – Self-education not included in the workflow

Nursing Implications • EBP Info. Bot provides free-text real time processing support • Evidence to support planning care at the point of care • Reduces information overload • Easy to use and navigate

Questions?